A method of estimating varietal stability for data of long-term trials